Column written for Interactions. © CACM, 2006. This is the author’s version of the work. It is posted here by permission of ACM for your personal use. It may be redistributed for non-commercial use only, provided this paragraph is included.
In my consulting activities, I often have to explain to companies that they are too logical, too rational. Human behavior seldom follows mathematical logic and reasoning. By the standards of engineers, human behavior can be illogical and irrational. From the standpoint of people, however, their behavior is quite sensible, dictated by the activity being performed, the environment and context, and their higher-level goals. To support real behavior we need activity-centered design.
Years ago, anthropologists Janet Dougherty and Charles Keller studied how blacksmiths organize their tools. Blacksmiths, they discovered, don't put all the hammers neatly away on the shelves, all together. No, when blacksmiths clean up at night, the hammer goes on the ground, right next to the anvil, and next to the tongs: all the tools are organized so that they are ready for the job, ready for use. In similar fashion, good carpenters, while working, keep nails near their hammers. In other words, good behavioral organization reflects human activity structure, not dictionary classification. Dougherty and Keller called this form of organization taskonomy.
Many of the design tools used by the Human-Centered Design community lead to well-structured, carefully organized designs, often using powerful card-sorting and hierarchical clustering algorithms to make similar things be located near one another. Call this the "Hardware store" organization. Hammers are in the hammer section where they are all logically arranged. Nails are in the nail section.
The hardware store organization is based upon a taxonomy: appropriate for libraries and for stores where the major problem is locating the desired item out of context. But note that some stores have learned to provide activity-centered organization in addition to their normal classification. Thus, smart food stores put potato chips and pretzels next to the beer. And some even put beer next to the diapers, so that when a shopper makes a late night, emergency trip to get more diapers, why there is the beer, temptingly convenient. Sensible, well-organized logical design would not support this real behavior.
Consider Walter Mossberg’s Wall Street Journal review of smart cellphones: his criticisms of the Treo 700w and the Motorola Q phones, both of which use the Microsoft OS, and his praise for the Treo 700p which uses PalmSource’s Palm OS. “The need to open menus and take other extra steps,” says Mossberg, “is endemic in the Windows Mobile software.”
In David Pogue’s New York Times review of the Motorola Q phone, the column headline says it all: “Lovely phone: Ugly software.” The software? Microsoft's Windows Mobile 5.0 OS:
“After you take a picture with the camera, what options would you want to be immediately available? Maybe Save, Send and Delete? Not on this phone. These options are all hiding in menus; activating Send, for example, requires four more button presses. (On the Treo: one.)”
The contrast between the Palm Treo 700w and 700p is especially powerful because both have essentially identical visual appearance, hardware, and physical buttons, the only difference being that the 700p uses the PalmSource OS, the 700w uses Microsoft's. The PalmSource OS readily wins at interaction design. Palm is organized around activities. Microsoft and Motorola are organized around action categories.
Why is this? I know that all companies involved here -- Microsoft, Palm, and Motorola -- have excellent user interface design teams. The answer is that the companies have, for whatever reason, followed different philosophies in their approach to interface design. Microsoft and Motorola went the taxonomic route, developing logical, well-structured interfaces for their Media Station, Windows, and Mobile OS. while Palm went the activity-centered, taskonomic route, with direct support for activities in the Palm OS.
The problem is related to the concerns voiced in my earlier essay, Human-Centered Design Considered Harmful. Many of the designs being produced by the HCD community are far too logical. They follow the hardware store approach to classification. This organization is well-suited for well-structured retrieval, but ill-suited for the direct support of an activity. We need both: nails to be stored with other nails when we are gathering them in preparation for an activity, but nails next to the hammer when we are in the midst of the activity.
This is why so much well-designed software fails. This is the problem with BMW’s original design for the iDrive. The iDrive provided a logical, sensible organization of the automobile’s controls and displays. But it failed to support activity patterns. The correct approach to the support of behavior is activity-based classification.
Is activity-centered design overthrowing all that we have learned about human-centered design? No, definitely not. I consider activity structure to be a refinement of HCD. Taxonomic structures are appropriate when there is no context, when suddenly needing some new piece of information or tool. That’s why this structure works well for libraries, stores, websites, and the program menu of an operating system. But once an activity has begun, then taskonomy is the way to go, where things used together are placed near one another, where any one item might be located logically within the taxonomic structure but also wherever behaviorally appropriate for the activities being supported.
The best solution is to provide both solutions: taxonomies and taskonomies. Some websites organize all their items logically and sensibly in a taxonomic structure, but once a particular item has been selected, taskonomic information appears. For example, if examining a pair of pants, the website might suggest shoes and shirts that match. Look at a printer and the website might suggest ink, paper, and other related accessories. Buy a book, and the website suggests other books on related topics, or sometimes, books purchased by other people who also bought the book under consideration. Such recommendations based upon past behavior are often superior to recommendations based upon logic.
Activity-centered design organizes according to usage: traditional human-centered design organizes according to topic, in isolation, outside the context of real, everyday use. Both are needed.
Dougherty, J., & Keller, C. (1985). Taskonomy: A practical approach to knowledge structures. In J. Dougherty (Ed.), Directions in cognitive anthropology (pp. 161-174). Urbana: University of Illinois Press.
Mossberg, W. (2006, January 5). A new Palm Treo uses Microsoft’s software, but it doesn’t beat 650. The Wall Street Journal.
Norman, D. A. (2005). Human-centered design considered harmful. Interactions, 12(4), 14-19.
Pogue, D. (2006, June 8). Motorola's Q: Lovely Phone; Ugly Software. New York Times.
Don Norman wears many hats, including co-founder of the Nielsen Norman group, Professor at Northwestern University, and author, his latest book being Emotional Design. Some of the material in this column is from his forthcoming book, The Design of Future Things (in preparation). He lives at www.jnd.org.